Healthcare information technology has great promise for improving care and reducing costs. However the safety of healthcare IT can itself be an issue. Optimal interaction design of electronic health records has proven difficult due to the great complexity, high-stakes and collaborative nature of healthcare work. The Institute of Medicine recently identified several areas of concern, such as display fragmentation. This project will study human-computer interaction and cognition in electronic health records, examining specific identified issues to establish basic science knowledge of how clinicians' (doctors and nurses) interaction and cognition are affected by system design, and how it could be improved to ensure safe systems. The areas examined include display fragmentation and cognitive load, interruption and nonlinear workflow, cognitive support for complex clinical reasoning including diagnosis and decision-making, and fit to task for different specialties and roles. We use a mixed method approach which incorporates both established methods such as usability testing and thinkaloud protocols, and novel methods including pupillometry to ascertain comparative cognitive load continuously during system use. A novel system which allows eliciting needs, rapid prototyping (in seconds) and testing of new configurations will be used to conduct controlled experiments of conventional and novel approaches. This work will elucidate how interaction design affects cognitive load and evaluate a possible approach for reducing it. It will also clarify the differing information needs and interaction requirements for different specialties and roles, and deepen our knowledge of how 'cognitive ergonomics' affects higher level cognition during clinical work. This can result in a set of specifications and design pattern available for incorporation into current systems.